Two-stage Convolutional Part Heatmap Regression for the 1st 3D Face Alignment in the Wild (3DFAW) Challenge
arXiv:1609.09545 · doi:10.1007/978-3-319-48881-3_43
Abstract
This paper describes our submission to the 1st 3D Face Alignment in the Wild (3DFAW) Challenge. Our method builds upon the idea of convolutional part heatmap regression [1], extending it for 3D face alignment. Our method decomposes the problem into two parts: (a) X,Y (2D) estimation and (b) Z (depth) estimation. At the first stage, our method estimates the X,Y coordinates of the facial landmarks by producing a set of 2D heatmaps, one for each landmark, using convolutional part heatmap regression. Then, these heatmaps, alongside the input RGB image, are used as input to a very deep subnetwork trained via residual learning for regressing the Z coordinate. Our method ranked 1st in the 3DFAW Challenge, surpassing the second best result by more than 22%.
arXiv commentWinner of 3D Face Alignment in the Wild (3DFAW) Challenge, ECCV 2016
References in corpus (1)
Cited by in corpus (4)
- Multi-Class Multi-Object Tracking using Changing Point Detection
- SADRNet: Self-Aligned Dual Face Regression Networks for Robust 3D Dense Face Alignment and Reconstruction
- Multi-view consensus CNN for 3D facial landmark placement
- Expression-preserving face frontalization improves visually assisted speech processing